COMPARATION OF CLASSIFICATION METHODS ON SENTIMENT ANALYSIS OF AHY ELECTABILITY BASED ON PUBLIC COMMENTS ON ONLINE NEWS MEDIA SITES
Windu Gata, Dewi Ayu Puspitawati, Syamsu Hidayat, Afri Yudha, Iskandar, Hasan Basri, Walim, Satria Wira Yudha, Saeful Bahri
Graduation School Master Degree Computer Science Sekolah Tinggi Manajemen Informatika dan Komputer Nusa Mandiri (STMIK Nusa Mandiri), Jl. Kramat 18 Jakarta Pusat. Indonesia
Abstract
In the run-up to the General Election in Indonesia, most political parties and prominent figures in the elections have started to prepare. Many of them deal with elections is one of them with their frequent public appearances to improve the electability party or leaders who supported. Activities they always decorate the news on online news media. Its real-time make news articles published online news online quickly get comments from readers. The comment is an opinion or expression of a people point of view of an object being preached, that used to see how the sentiments raised by the readers. The result shows that SVM method have the accuracy of 76.09% and AUC 0.848 value with Good Classification diagnosis, while for NB method it produces 68.21% accuracy and AUC value 0.672 with Poor Classification diagnosis. However, after optimizing the PSO method the accuracy of the SVM method became 78.40% and the value of AUC 0.850 with Good Classification diagnosis, while for NB method became 74.98% and AUC 0.708 with Fair Classification diagnosis.
Keywords: data mining, AHY, indonesian text mining, naive bayes, svm
Topic: Informatic and Information System